Political Economy of Institutions and Development. Lectures 3-5: Introduction to Dynamic Voting and Constitutions

Transcription

1 Political Economy of Institutions and Development. Lectures 3-5: Introduction to Dynamic Voting and Constitutions Daron Acemoglu MIT March 8, 13 and 15. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

2 Dynamic Voting Models and Constitutions Introduction Introduction Markov Perfect Equilibria different from myopic rules because they take into account the effect of current votes on future political decisions. These issues are more salient and important when current political decisions affect the distribution of political power in the future. The set of issues that arise here are very similar to those that will be central when we think about endogenous institutions. Thus useful to start considering more general dynamic voting models. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

3 Dynamic Voting Models and Constitutions Introduction Why Worry about the Dynamics of Political Power? Why does political equilibrium lead to ineffi ciency? Why doesn t even the most corrupt and kleptocratic dictator just choose economically effi cient actions and then redistribute things towards himself? In static models, as we have seen, political equilibria are often Pareto effi cient (though often ineffi cient in other ways). In dynamic models, there is a new reason why ineffi ciency will arise: the political losers effect. If you do the right thing, this may reduce your political power and your future rents. To study these issues, we need dynamic models with endogenous distribution of political power. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

4 Examples Dynamic Voting Dynamic Voting in Clubs Let us start with a model due to Robert s (1999). Voting directly over club size (utilities directly from club size). Relatively parsimonious model, but it gets quickly complicated. Nevertheless, some important insights can be obtained. We will see both later in the lecture and when we study endogenous institutions later in the class how similar insights arise in different settings. Key issue: what type of structure we should impose on dynamic models so that they are tractable, while capturing real-world relevant phenomena? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

5 Examples Specifics of Roberts s Model Dynamic Voting An economy consisting of a finite group N = {1, 2..., n}. To make the model tractable, it is assumed that there is an actual seniority system whereby if the voting population is of size x, it includes individuals {1, 2,..., x}, i.e., lower index individuals are always included before higher index individuals. Let the set of potential clubs be denoted by S (these are sets of the form {1}, {1, 2}, etc.). Let us denote the size of the voting population at time t by x t and assume that the instantaneous utility of individual ξ when the size of the (voting) club is x is given by w ξ (x). In terms of more micro models, this instantaneous utility function incorporates what the utility of individual ξ will be when tax policies are determined by a club of x individuals. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

6 Examples Dynamic Voting Utility Given this instantaneous utility function, the expected utility of individual ξ at time t = 0 is given by: where 0 < β < 1 is the discount factor. Key assumption: U 0 = E 0 t=0 β t w ξ (x t ), (1) Assumption (Strict Increasing Differences) For all x > x, ξ > ξ, we have w ξ (x) w ξ (x ) > w ξ (x) w ξ (x ). Slight variant on single crossing. Higher ranked individuals included later in the club than lower-ranked individuals, but also have preference towards larger groups. This would make sense, for example, when we think of larger franchises as leading to higher taxes, and higher taxes being more damaging to richer individuals. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

7 Constitutional Choice Examples Constitutional Choice autocracy limited franchise full democracy Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

8 Examples Constitutional Choice Constitutional Choice Simple Example Three states: absolutism a, constitutional monarchy c, full democracy d Two agents: elite E, middle class M E rules in a, M rules in c and d. w E (d) < w E (a) < w E (c) w M (a) < w M (c) < w M (d) Myopic elite: starting from a, move to c Farsighted elite (high discount factor): stay in a as moving to c will lead to M moving to d But very different insights when there are stochastic elements and intermediate discount factors. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

9 States and Utilities Main Tools and Ideas Main Tools and Ideas More formally, society starts period in state (e.g., size of club, constitution, policy) s t 1 and decides on (feasible) s t Individual i in period t gets instantaneous utility w i (s t ) Strict increasing differences: For any agents i, j N such that i > j, w i (s) w j (s) is increasing in s This could be weakened to weak increasing differences for some results. In addition, possibly transition cost c i (s t, s t 1 ). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

10 Transition Mapping Main Tools and Ideas Main Tools and Ideas Let us consider Markov transition rules for analyzing how the state changes over time. A Markov transition rule is denoted by φ such that φ : S S. A transition rule is useful because it defines the path of the state s recursively such that for all t, i.e., s t+1 = φ(s t ). Why Markov? If there is an s such that s = φ(s ), then s is a steady state of the system. We will consider both deterministic and stochastic transition rules φ( ). But for now, useful to think of it as non-stochastic. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

11 Main Tools and Ideas Recursive Representation Main Tools and Ideas Value function (conditioned on transition mapping φ): V φ i (s) = w i (s) + Recursively k=1 V φ i (s) = w i (s) + β [ ) ( )] β k w i (φ k (s) c i φ k 1 (s), φ k (s). [ ] V φ i (φ(s)) c i (s, φ (s)), or V φ i (s) = w i (s) βc i (s, φ (s)) + βv φ i (φ(s)) Also define continuation value inclusive of transition costs: V φ i (s x) = V φ i (s) c i (x, s) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

12 Main Tools and Ideas Main Tools and Ideas Recursive Representation (continued) In the stochastic case: V φ E,i (s) = w E,i (s) + β E q ( E, E ) [ ] V φ E,i (φ E (s)) c E,i (s, φ E (s)) E where E denotes different environments with different payoffs, transition costs or political processes, and q (E, E ) denotes transition probabilities. Also: V φ E,i (s x) = V φ E,i (s)) c E,i (x, s). Key observation: If w satisfies (strict) increasing differences, than so does V. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

13 Main Tools and Ideas Main Tools and Ideas Markov Voting Equilibrium φ = {φ E : S S} is a Markov Voting Equilibrium if for any x, y S, {i N : V φ E,i (y x) > V φ E,i (φ E (x) x) } {i N : V φ E,i (φ E (x) x) V φ E,i (x) } / W E,x W E,x The first is ensures that there isn t another state transition to which would gather suffi cient support. Analogy to core. The second one ensures that there is a winning coalition supporting the transition. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

14 Roadmap Main Tools and Ideas Main Tools and Ideas We now study some special cases, then returning to the general framework so far outlined. A (finite) game of political eliminations. Characterization for the general model without stochastic elements and with β close to 1. Applications. Characterization for the general model with stochastic elements and arbitrary discount factor β. Applications. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

15 Voting over Coalitions Voting over Coalitions Voting Over Coalitions Another obvious example of dynamic voting with changing constituencies. Model based on Acemoglu, Egorov and Sonin (2008). A coalition, which will determine the distribution of a pie (more generally payoffs), both over its own membership. Possibility of future votes shaping the stability of current clubs illustrated more clearly. Motivation: 1 the three-player divide the dollar game. 2 eliminations in the Soviet Politburo. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

16 Voting over Coalitions Voting over Coalitions Political Game Let I denote the collection of all individuals, which is assumed to be finite. The non-empty subsets of I are coalitions and the set of coalitions is denoted by C. For any X I, C X denotes the set of coalitions that are subsets of X and X is the number of members in X. In each period there is a designated ruling coalition, which can change over time. The game starts with ruling coalition N, and eventually the ultimate ruling coalition (URC) forms. When the URC is X, then player i obtains baseline utility w i (X ) R. w ( ) {w i ( )} i I. Important assumption: game of non-transferable utility. Why? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

17 Voting over Coalitions Voting over Coalitions Political Power So far, our focus has been on democratic situations. One person one vote. Now allow differential powers across individuals. Power mapping to: γ : I R ++, γ i γ (i): political power of individual i I and γ X i X γ i political power of coalition X. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

18 Voting over Coalitions Voting over Coalitions Winning Coalitions Coalition Y X is winning within coalition X if and only if γ Y > αγ X, where α [1/2, 1) is a (weighted) supermajority rule (α = 1/2 corresponds to majority rule). Let us write: Y W X for Y X winning within X. Since α 1/2, if Y, Z W X, then Y Z =. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

19 Payoffs Voting over Coalitions Voting over Coalitions Assumption: Let i I and X, Y C. Then: (1) If i X and i / Y, then w i (X ) > w i (Y ) [i.e., each player prefers to be part of the URC]. (2) For i X and i Y, w i (X ) > w i (Y ) γ i /γ X > γ i /γ Y ( γ X < γ Y ) [i.e., for any two URCs that he is part of, each player prefers the one where his relative power is greater]. (3) If i / X and i / Y, then w i (X ) = w i (Y ) w i [i.e., a player is indifferent between URCs he is not part of]. Interpretation. Example: w i (X ) = γ { X {i} γi /γ = X if i X γ X 0 if i / X. (2) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

20 Extensive-Form Game Voting over Coalitions Voting over Coalitions Choose ε > 0 arbitrarily small. Then, the extensive form of the game Γ = (N, γ N, w ( ), α) is as follows. Each stage j of the game starts with some ruling coalition N j (at the beginning of the game N 0 = N). Then: 1. Nature randomly picks agenda setter a j,q N j for q = [Agenda-setting step] Agenda setter a j,q makes proposal P j,q C Nj, which is a subcoalition of N j such that a j,q P j,q (for simplicity, we assume that a player cannot propose to eliminate himself). 3. [Voting step] Players in P j,q vote sequentially over the proposal. More specifically, Nature randomly chooses the first voter, v j,q,1, who then casts his vote vote ṽ (v j,q,1 ) {ỹ, ñ} (Yes or No), then Nature chooses the second voter v j,q,2 = v j,q,1 etc. After all P j,q players have voted, the game proceeds to step 4 if players who supported the proposal form a winning coalition within N j (i.e., if {i P j,q : ṽ (i) = ỹ} W Nj ), and otherwise it proceeds to step 5. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

21 Voting over Coalitions Extensive-Form Game (continued) Voting over Coalitions 4. If P j,q = N j, then the game proceeds to step 6. Otherwise, players from N j \ P j,q are eliminated and the game proceeds to step 1 with N j+1 = P j,q (and j increases by 1 as a new transition has taken place). 5. If q < N j, then next agenda setter a j,q+1 N j is randomly picked by Nature among members of N j who have not yet proposed at this stage (so a j,q+1 = a j,r for 1 r q), and the game proceeds to step 2 (with q increased by 1). If q = N j, the game proceeds to step N j becomes the ultimate ruling coalition. Each player i N receives total payoff U i = w i (N j ) ε 1 k j I {i Nk }, (3) where I { } is the indicator function taking the value of 0 or 1. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

22 Voting over Coalitions Voting over Coalitions Discussion Natural game of sequential choice of coalitions. ε introduced for technical reasons (otherwise, indifferences lead to uninteresting transitions). Important assumption: players eliminated have no say in the future. Stark representation of changing constituencies, but not a good approximation to democratic decision-making. More reminiscent to dealmaking in autocracies or coalition formation in nondemocracies. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

23 Voting over Coalitions Axiomatic Analysis Axiomatic Analysis Games of coalition formation have noncooperative and cooperative features. Ideally, the two perspectives give congruent results. Key idea in the extensive-form game: players will not support a coalition that will later eliminate themselves. stability Let us first capture this notion using an axiomatic approach. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

24 Voting over Coalitions Axiomatic Analysis Axioms Consider the set of games (N, γ N, w ( ), α). Holding γ, w and α fixed, consider the correspondence φ : C C defined by φ (N) = Φ (N, γ N, w, α) for any N C. Axioms on φ. Axiom 1 (Inclusion) For any X C, φ (X ) = and if Y φ (X ), then Y X. Axiom 2 (Power) For any X C, Y φ (X ) only if Y W X. Axiom 3 (Self-Enforcement) For any X C, Y φ (X ) only if Y φ (Y ). Axiom 4 (Rationality) For any X C, for any Y φ (X ) and for any Z X such that Z W X and Z φ (Z ), we have that Z / φ (X ) γ Y < γ Z. Interpretation. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

25 Voting over Coalitions Self-Enforcing Coalitions Axiomatic Analysis Motivated by the self-enforcement axiom: Definition Coalition X P (I) is self-enforcing if X φ (X ). Assumption: The power mapping γ is generic in the sense that if for any X, Y C, γ X = γ Y implies X = Y. We also say that coalition N is generic or that numbers {γ i } i N are generic if mapping γ N is generic. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

26 Main Axiomatic Result Voting over Coalitions Axiomatic Analysis Theorem Fix I, γ, w ( ) and α [1/2, 1). Then: 1. There exists a unique mapping φ that satisfies Axioms 1 4. Moreover, when γ is generic, φ is single-valued. 2. This mapping φ may be obtained by the following inductive procedure. For any k N, let C k = {X C : X = k}. Clearly, C = k N C k. If X C 1, then let φ (X ) = {X }. If φ (Z ) has been defined for all Z C n for all n < k, then define φ (X ) for X C k as φ (X ) = argmin γ A, and (4) A M(X ) {X } M (X ) = {Z C X \ {X } : Z W X and Z φ (Z )}. (5) Proceeding inductively φ (X ) is defined for all X C. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

27 Voting over Coalitions Axiomatic Analysis Intuition For each X, (5) defines M (X ) as the set of proper subcoalitions which are both winning and self-enforcing. Equation (4) then picks the coalitions in M (X ) that have the least power. When there are no proper winning and self-enforcing subcoalitions, M (X ) is empty and X becomes the URC), which is captured by (4). What does this mean? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

28 Voting over Coalitions Axiomatic Analysis Implication Corollary Coalition N is self-enforcing, that is, N φ (N), if and only if there exists no coalition X N, X = N, that is winning within N and self-enforcing. Moreover, if N is self-enforcing, then φ (N) = {N}. Main implication: a coalition that includes a winning and self-enforcing subcoalition cannot be self-enforcing. This captures the notion that the stability of smaller coalitions undermines stability of larger ones. Application: coalition formation among three players with approximately equal powers. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

29 Noncooperative Game Voting over Coalitions Equilibrium Characterization Theorem Suppose that φ (N) satisfies Axioms 1-4 (cfr. (4) in the axiomatic analysis). Then, for any K φ (N), there exists a pure strategy profile σ K that is an SPE and leads to URC K in at most one transition. In this equilibrium player i N receives payoff U i = w i (K ) εi {i K } I {N =K }. (6) This equilibrium payoff does not depend on the random moves by Nature. Thus equivalence between cooperative and noncooperative approaches. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

30 Intuition Voting over Coalitions Equilibrium Characterization Suppose each player anticipates members of a self-enforcing ruling coalition to play a strategy profile such that they will turn down any offers other than K and they will accept K; then, it is in the interest of all the players in K to play such a strategy for any history. This follows immediately because by the definition of the set φ (N), because for any deviation to be profitable, the URC that emerges after such deviation must be either not self-enforcing or not winning. But the the first option is ruled out by induction while a deviation to a non-winning URC will be blocked by suffi ciently many players. The payoff in (6) is also intuitive. Each player receives his baseline payoff w i (K ) resulting from URC K and then incurs the cost ε if he is part of K and if the initial coalition N is not equal to K (because in this latter case, there will be one transition). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

31 Voting over Coalitions Stronger Results Under Genericity Equilibrium Characterization Theorem Suppose the genericity Assumption holds and suppose φ (N) = K. Then any (pure or mixed strategy) SPE results in K as the URC. The payoff player i N receives in this equilibrium is given by (6). Intuition. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

32 Voting over Coalitions The Structure of Ruling Coalitions Characterization Equilibrium characterize simply by a set of recursive equations. What are the implications of equilibrium coalition formation Let us impose one more assumption Assumption: For no X, Y C such that X Y the equality γ Y = αγ X is satisfied. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

33 Voting over Coalitions Continuity of Ruling Coalitions The Structure of Ruling Coalitions Proposition Consider Γ = (N, γ, w ( ), α) with α [1/2, 1). Then: 1. There exists δ > 0 such that if γ : N R ++ lies within δ-neighborhood of γ, then Φ (N, γ, w, α) = Φ (N, γ, w, α). 2. There exists δ > 0 such that if α [1/2, 1) satisfies α α < δ, then Φ (N, γ, w, α) = Φ (N, γ, w, α ). 3. Let N = N 1 N 2 with N 1 and N 2 disjoint. Then, there exists δ > 0 such that for all N 2 such that γ N2 < δ, φ (N 1 ) = φ (N 1 N 2 ). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

34 Voting over Coalitions Fragility of Self-Enforcing Coalitions The Structure of Ruling Coalitions Proposition Suppose α = 1/2 and fix a power mapping γ : I R ++. Then: 1. If coalitions X and Y such that X Y = are both self-enforcing, then coalition X Y is not self-enforcing. 2. If X is a self-enforcing coalition, then X {i} for i / X and X \ {i} for i X are not self-enforcing. Implication: under majority rule α = 1/2, the addition or the elimination of a single agent from a self-enforcing coalitions makes this coalition no longer self-enforcing. Why? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

35 Voting over Coalitions The Structure of Ruling Coalitions Size of Ruling Coalitions I Proposition Consider Γ = (N, γ, w ( ), α). 1. Suppose α = 1/2, then for any n and m such that 1 m n, m = 2, there exists a set of players N, N = n, and a generic mapping of powers γ such that φ (N) = m. In particular, for any m = 2 there exists a self-enforcing ruling coalition of size m. However, there is no self-enforcing coalition of size Suppose that α > 1/2, then for any n and m such that 1 m n,there exists a set of players N, N = n, and a generic mapping of powers γ such that φ (N) = m. Therefore, one can say relatively little about the size and composition of URCs without specifying the power distribution within the society further (except that when α = 1/2, coalitions of size 2 are not self-enforcing). But this is because no discipline on the distribution of powers. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

36 Voting over Coalitions Size of Ruling Coalitions II The Structure of Ruling Coalitions Proposition Consider Γ = (N, γ, w ( ), α) with α [1/2, 1). Suppose that there exists δ > 0 such that max i,j N { γi /γ j } < 1 + δ. Then: 1. When α = 1/2, any ruling coalition must have size k m = 2 m 1 for some m Z, and moreover, φ (N) = N if and only if N = k m for k m = 2 m When α [1/2, 1), φ (N) = N if and only if N = k m,α where k 1,α = 1 and k m,α = k m 1,α /α + 1 for m > 1, where z denotes the integer part of z. When powers are approximately equal, the size of the URC is determined tightly. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

37 Rules and Coalitions Voting over Coalitions The Structure of Ruling Coalitions Should an increase in α raise the size of the URC? Should an individual always gain from an increase in his power? Intuitive, but the answers are no and no. Proposition 1. An increase in α may reduce the size of the ruling coalition. That is, there exists a society N, a power mapping γ and α, α [1/2, 1), such that α > α but for all X Φ (N, γ, w, α) and X Φ (N, γ, w, α ), X > X and γ X > γ X. 2. There exist a society N, α [1/2, 1), two mappings γ, γ : N R ++ satisfying γ i = γ i for all i = j, γ j < γ j such that j Φ (N, γ, w, α), but j / Φ (N, γ, w, α). Moreover, this result applies even when j is the most powerful player in both cases, i.e. γ i = γ i < γ j < γ j for all i = j. Why? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

38 Voting over Coalitions Power and Ruling Coalitions The Structure of Ruling Coalitions When will the most powerful individual be part of the ruling coalition? Proposition Consider the game Γ (N, γ, w ( ), α) with α [1/2, 1), and suppose that γ 1,..., γ N is an increasing sequence. If ( ) N 1 N 1 γ N α j=2 γ j / (1 α), α j=1 γ j / (1 α), then either coalition N is self-enforcing or the most powerful individual, N, is not a part of the URC. Intuition? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

39 Voting over Coalitions Conclusions Conclusions Once dynamic voting also affects the distribution of political power, richer set of issues arise. Endogeneity of constituencies is both practically relevant and related to endogenous institutions. Ensuring equilibria in situations of dynamic voting harder, but often we can put economically interesting structure to ensure equilibria (once we know what we are trying to model). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

40 Constitutions Introduction Introduction Why do constitutions matter? What is written in constitutions seems to matter, but constitutions can be disobeyed and rewritten. How do we think about the role of constitutions? Different approaches. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

41 Philosophical Constitutions Introduction What is written on paper should not matter: because whatever is written down could have been expected even when it was not written down a constitution is as good as the force behind it But this perspective may not be too useful in studying how constitutions are written in practice, why they persist and why and when they matter. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

42 Constitutions General Approach Dynamics and Stability: a More General Approach A more general approach towards stability and change in social arrangements (political regimes, constitutions, coalitions, clubs, firms) without giving up existence. Essential ingredients: Payoffs: different arrangements imply different payoffs Power: different arrangements reallocate political or decision-making power In this light, we need to study: Change: which arrangements will be changed by force or reform Stability: which arrangements will resist change Are there any general insights? Strategy: Formulate a general dynamic framework to investigate the interplay of these two factors in a relatively detail-free manner. Details useful to go beyond general insights. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

43 Constitutions General Approach Simple Example Consider a simple extension of franchise story Three states: absolutism a, constitutional monarchy c, full democracy d Two agents: elite E, middle class M E rules in a, M rules in c and d. w E (d) < w E (a) < w E (c) w M (a) < w M (c) < w M (d) Myopic elite: starting from a, move to c Farsighted elite: stay in a: move to c will lead to M moving to d. Same example to illustrate resistance against socially beneficial reform. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

44 Constitutions General Approach Naïve and Dynamic Insights Naïve insight: a social arrangement will emerge and persist if a suffi ciently powerful group prefers it to alternatives. Simple example illustrates: power to change towards a more preferred outcome is not enough to implement change because of further dynamics Social arrangements might be stable even if there are powerful groups that prefer change in the short run. Key: social arrangements change the distribution of political power (decision-making capacity). Dynamic decision-making: future changes also matter (especially if discounting is limited) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

45 Constitutions General Approach Applications Key motivation: changes in constitutions and political regimes. Extension of franchise (Acemoglu and Robinson 2000, 2006, Lizzeri and Persico 2004) Members of a club decide whether to admit additional members by majority voting (Roberts 1999) Society decides by voting, what degree of (super)majority is needed to start a reform (Barbera and Jackson 2005) EU members decide whether to admit new countries to the union (Alesina, Angeloni, and Etro 2005) Inhabitants of a jurisdiction determine migration policy (Jehiel and Scotchmer 2005) Participant of (civil) war decides whether to make concessions to another party (Fearon 1998, Schwarz and Sonin 2008) Dynamic political coalition formation: Junta (or Politburo) members decide whether to eliminate some of them politically or physically (Acemoglu, Egorov, and Sonin 2008) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

46 Model and Approach Constitutions General Approach Model: Finite number of individuals. Finite number of states (characterized by economic relations and political regimes) Payoff functions determine instantaneous utility of each individual as a function of state Political rules determine the distribution of political power and protocols for decision-making within each state. A dynamic game where politically powerful groups can induce a transition from one state to another at any date. Question: what is the dynamically stable state as a function of the initial state? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

47 Constitutions General Approach Main Results of General Framework An axiomatic characterization of outcome mappings corresponding to dynamic game (based on a simple stability axiom incorporating the notion of forward-looking decisions). Equivalence between the MPE of the dynamic game (with high discount factor) and the axiomatic characterization Full characterization: recursive and simple Under slightly stronger conditions, the stable outcome (dynamically stable state) is unique given the initial state but depends on the initial state Model general enough to nest specific examples in the literature. In particular, main theorems directly applicable to situations in which states can be ordered and static payoffs satisfy single crossing or single peakedness. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

48 Simple Implications Constitutions General Approach A particular social arrangement is made stable by the instability of alternative arrangements that are preferred by suffi ciently many members of the society. stability of a constitution does not require absence of powerful groups opposing it, but the absence of an alternative stable constitution favored by powerful groups. Effi ciency-enhancing changes are often resisted because of further social changes that they will engender. Pareto ineffi cient social arrangements often emerge as stable outcomes. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

49 Constitutions Model Model: Basics Finite set of individuals I ( I total) Set of coalitions C (non-empty subsets X I) Each individual maximizes discounted sum of playoffs with discount factor β [0, 1). Finite set of states S ( S total) Discrete time t 1 State s t is determined in period t; s 0 is given Each state s S is characterized by Payoff w i (s) of individual i I (normalize w i (s) > 0) Set of winning coalitions W s C capable of implementing a change Protocol π s (k), 1 k K s : sequence of agenda-setters or proposals (π s (k) I S) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

50 Constitutions Model Winning Coalitions Assumption (Winning Coalitions) For any state s S, W s C satisfies two properties: (a) If X, Y C, X Y, and X W s then Y W s. (b) If X, Y W s, then X Y =. (a) says that a superset of a winning coalition is winning in each state (b) says that there are no two disjoint winning coalitions in any state W s = is allowed (exogenously stable state) Example: Three players 1, 2, 3 W s = {{1}, {1, 2}, {1, 3}, {1, 2, 3}} is valid (1 is dictator) W s = {{1, 2}, {1, 3}, {2, 3}, {1, 2, 3}} is valid (majority voting) W s = {{1}, {2, 3}} is not valid (both properties are violated) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

51 Dynamic Game Constitutions Model 1 Period t begins with state s t 1 from the previous period. 2 For k = 1,..., K st 1, the kth proposal P k,t is determined as follows. If π st 1 (k) S, then P k,t = π st 1 (k). If π st 1 (k) I, then player π st 1 (k) chooses P k,t S. 3 If P k,t = s t 1, each player votes (sequentially) yes (for P k,t ) or no (for s t 1 ). Let Y k,t denote the set of players who voted yes. If Y k,t W t 1, then P k,t is accepted, otherwise it is rejected. 4 If P k,t is accepted, then s t = P k,t. If P k,t is rejected, then the game moves to step 2 with k k + 1 if k < K st 1. If k = K st 1, s t = s t 1. 5 At the end of each period (once s t is determined), each player receives instantaneous utility u i (t): { wi (s) if s u i (t) = t = s t 1 = s 0 if s t = s t 1 Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

52 Constitutions Key Notation and Concepts Model Define binary relations: states x and y are payoff-equivalent x y i I : w x (i) = w y (i) y is weakly preferred to x in z y z x {i I : w y (i) w x (i)} W z y is strictly preferred to x in z y z x {i I : w y (i) > w x (i)} W z Notice that these binary relations are not simply preference relations they encode information about preferences and political power. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

53 Constitutions Model Preferences and Acyclicity Assumption (Payoffs) Payoff functions {w i ( )} i I satisfy: (a) For any sequence of states s 1,..., s k in S, s j+1 sj s j for all 1 j k 1 = s 1 sk s k. (b) For any sequence of states s, s 1,..., s k in S such that s j s l (for any 1 j < l k) and s j s s (for any 1 j k) s j+1 s s j for all 1 j < k 1 = s 1 s s k. Moreover, if for x, y, s in S, we have x s s and y s s, then y s x. (a) rules out cycles of the form y z z, x y y, z x x (b) rules out cycles for and also an additional condition that is weaker than transitivity within states, i.e., x s s, y s s, then y s x, whereas transitivity would require x s s, s s y, then x s y, which implies our condition, but is much stronger. Alternative (with equivalent results): voting yes has a small cost. These assumptions rule out Condorcet-type cycles emerge. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

54 Constitutions Model Preferences and Acyclicity (continued) We will also strengthen our results under: Assumption (Comparability) For x, y, z S such that x z z, y z z, and x y, either y z x or x z y. This condition suffi cient (and necessary ) for uniqueness. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

55 Constitutions Axiomatic Characterization Approach and Motivation Key economic insight: with suffi ciently forward-looking behavior, an individual should not wish to transition to a state that will ultimately lead to another lower utility state. Characterize the set of allocations that are consistent with this insight without specifying the details of the dynamic game. Introduce three simple and intuitive axioms. Characterize set of mappings Φ such that for any φ Φ, φ : S S satisfies these axioms and assigns an axiomatically stable state s S to each initial state s 0 S (i.e., φ (s) = s S loosely corresponding to s t = s for all t T for some T ). Interesting in its own right, but the main utility of this axiomatic approach is as an input into the characterization of the (two-strategy) MPE of the dynamic game. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

56 Constitutions Axiomatic Characterization Axiom 1 (Desirability) If x, y S are such that y = φ (x), then either y = x or y x x. A winning coalition can always stay in x (even a blocking coalition can) A winning coalition can move to y If there is a transition to y, a winning coalition must have voted for that Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

57 Constitutions Axiomatic Characterization Axiom 2 (Stability) If x, y S are such that y = φ (x), then y = φ (y). Holds by definition of φ ( ): T : s t = φ (s) for all t T ; when φ (s) is reached, there are no more transitions If y were unstable (y = φ (y)), then why not move to φ (y) instead of y Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

58 Constitutions Axiomatic Characterization Axiom 3 (Rationality) If x, y, z S are such that z x x, z = φ (z), and z x y, then y = φ (x). A winning coalition can move to y and to z A winning coalition can stay in x When will a transition to y be blocked? If there is another z preferred by some winning coalition If this z is also preferred to x by some winning coalition (so blocking y will lead to z, not to x) If transition to z is credible in the sense that this will not lead to some other state in perpetuity Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

59 Constitutions Axiomatic Characterization Stable States State s S is φ-stable if φ (s) = s for φ Φ Set of φ-stable states: D φ = {s S: φ (s) = s for φ Φ} We will show that if φ 1 and φ 2 satisfy the Axioms, then D φ1 = D φ2 = D Even if φ is non-unique, notion of stable state is well-defined But φ 1 (s) and φ 2 (s) may be different elements of D Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

60 Constitutions Main Axiomatic Theorem Axiomatic Characterization of Stable States Theorem Suppose Assumptions on Winning Coalitions and Payoffs hold. Then: 1 There exists mapping φ satisfying Axioms This mapping φ may be obtained through a recursive procedure (next slide) 3 For any two mappings φ 1 and φ 2 that satisfy Axioms 1 3 the the sets of stable states of these mappings coincide (i.e., D φ1 = D φ2 = D). 4 If, in addition, the Comparability Assumption holds, then the mapping that satisfies Axioms 1 3 is payoff-unique in the sense that for any two mappings φ 1 and φ 2 that satisfy Axioms 1 3 and for any s S, φ 1 (s) φ 2 (s). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

61 Recursive Procedure Constitutions Main Axiomatic Theorem Theorem (continued) Any φ that satisfies Axioms 1 3 can be recursively } computed as follows. Construct the sequence of states {µ 1,..., µ S with the property that if for any l (j, S ], µ l µj µ j. Let µ 1 S be such that φ (µ 1 ) = µ 1. For k = 2,..., S, let M k = { s { µ 1,..., µ k 1 } : s µk µ k and φ (s) = s }. Define, for k = 2,..., S, { µ φ (µ k ) = k if M k = z M k : x M k with x µk z if M k =. (If there exist more than one s M k : z M k with z µk any of these; this corresponds to multiple φ functions). s, we pick Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

62 Constitutions Example Extension of Franchise Example Get back to the simple extension of franchise story Three states: absolutism a, constitutional monarchy c, full democracy d Two agents: elite E, middle class M w E (d) < w E (a) < w E (c) w M (a) < w M (c) < w M (d) W a = {{E }, {E, M}}, W c = {{M}, {E, M}}, W d = {{M}, {E, M}} Then: φ (d) = d, φ (c) = d, therefore, φ (a) = a Indeed, c is unstable, and among a and d player E, who is part of any winning coalition, prefers a Intuitively, if limited franchise immediately leads to full democracy, elite will not undertake it Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

63 Constitutions Example Example (continued) Assume W c = {{E, M}} instead of W c = {{M}, {E, M}} Then: φ (d) = d, φ (c) = c, and, φ (a) = c a became unstable because c became stable Now assume W a = W c = W d = {{E, M}} and w E (a) < w E (d) < w E (c) w M (a) < w M (c) < w M (d) a is disliked by everyone, but otherwise preferences differ Then: φ (d) = d, φ (c) = c, and φ (a) may be c or d In any case, D = {c, d} is the same Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

64 Constitutions Noncooperative Characterization Back to Dynamic Game Assumption (Agenda-Setting and Proposals) For every state s S, one (or both) of the following two conditions is satisfied: (a) For any state q S \ {s}, there is an element k : 1 k K s of sequence π s such that π s (k) = q. (b) For any player i I there is an element k : 1 k K s of sequence π s such that π s (k) = i. Exogenous agenda, sequence of agenda-setters, or mixture. This assumption ensures that all proposals will be considered (or all agenda-setters will have a chance to propose) Definition (Dynamically Stable States) State s S is a dynamically stable state if there exist a protocol {π s } s S, a MPE strategy profile σ (for a game starting with initial state s 0 ) and T <, such that in MPE s t = s for all t T. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

65 Constitutions Noncooperative Characterization Noncooperative Characterization Theorem (Noncooperative Characterization) Suppose Assumptions on Winning Coalitions and Payoffs hold. Then there exists β 0 [0, 1) such that for all β β 0, the following results hold. 1 For any mapping φ satisfying Axioms 1 3 there is a protocol {π s } s S and a MPE σ of the game such that s t = φ (s 0 ) for any t 1; that is, the game reaches φ (s 0 ) after one period and stays in this state thereafter. Therefore, s = φ (s 0 ) is a dynamically stable state. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

66 Constitutions Noncooperative Characterization Noncooperative Characterization (continued) Theorem... Moreover: 2. For any protocol {π s } s S there exists a MPE in pure strategies. Any such MPE σ has the property that for any initial state s 0 S, it reaches some state, s by t = 1 and thus for t 1, s t = s. Moreover, there exists mapping φ : S S that satisfies Axioms 1 3 such that s = φ (s 0 ). Therefore, all dynamically stable states are axiomatically stable. 3. If, in addition, Assumption (Comparability) holds, then the MPE is essentially unique in the sense that for any protocol {π s } s S, any MPE strategy profile in pure strategies σ induces s t φ (s 0 ) for all t 1, where φ satisfies Axioms 1 3. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

67 Constitutions Myopic Stability and Ineffi ciency Dynamic vs. Myopic Stability Definition State s m S is myopically stable if there does not exist s S with s s m s m. Corollary 1 State s S is a (dynamically and axiomatically) stable state only if for any s S with s s s, and any φ satisfying Axioms 1 3, s = φ (s ). 2 A myopically stable state s m is a stable state. 3 A stable state s is not necessarily myopically stable. E.g., state a in extension of franchise story Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

68 Constitutions Myopic Stability and Ineffi ciency Ineffi ciency Definition (Inffi ciency) State s S is (strictly) Pareto ineffi cient if there exists s S such that w i (s ) > w i (s) for all i I. State s S is (strictly) winning coalition ineffi cient if there exists a winning coalition W s I in s and s S such that w i (s ) > w i (s) for all i W s. Clearly, if a state s is Pareto ineffi cient, it is winning coalition ineffi cient, but not vice versa. Corollary 1 A stable state s S can be (strictly) winning coalition ineffi cient and Pareto ineffi cient. 2 Whenever s is not myopically stable, it is winning coalition ineffi cient. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

69 Constitutions Ordered State Space Applying the Theorems in Ordered Spaces The characterization theorems provided so far are easily applicable in a wide variety of settings. In particular, if the set of states is ordered and static preferences satisfy single crossing or single peakedness, all the results provided so far can be applied directly. Here, for simplicity, suppose that I R and S R (more generally, other orders on the set of individuals and the set of states would work as well) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

70 Constitutions Ordered State Space Single Crossing and Single Peakedness Definition Take set of individuals I R, set of states S R, and payoff functions w ( ). Then, single crossing condition holds if whenever for any i, j I and x, y S such that i < j and x < y, w i (y) > w i (x) implies w j (y) > w j (x) and w j (y) < w j (x) implies w i (y) < w i (x). Definition Take set of individuals I R, set of states S R, and payoff functions w ( ). Then, single-peaked preferences assumption holds if for any i I there exists state x such that for any y, z S, if y < z x or x z > y, then w i (y) w i (z). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

71 Constitutions Ordered State Space Generalizations of Majority Rule and Median Voter Definition Take set of individuals I R, state s S, and set of winning coalitions W s that satisfies Assumption on Winning Coalitions. Player i I is called quasi-median voter (in state s) if i X for any X W s such that X = {j I : a j b} for some a, b R. That is, quasi-median voter is a player who belongs to any connected winning coalition. Quasi-median voters: simple majority 5/6 supermajority Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

72 Constitutions Ordered State Space Generalizations of Majority Rule and Median Voter (continued) Denote the set of quasi-median voters in state s by M s (it will be nonempty) Definition Take set of individuals I R, set of states S R. The sets of winning coalitions {W s } s S has monotonic quasi-median voter property if for each x, y S satisfying x < y there exist i M x, j M y such that i j Robert s model; ok also ok not ok Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

73 Constitutions Ordered State Space A Weak Genericity Assumption Let us say that preferences w ( ), given the set of winning coalitions {W s } s S, are generic if for all x, y, z S, x z y implies x z y or x y. This is (much) weaker than the comparability assumption used for uniqueness above. In particular, it holds generically. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

74 Constitutions Ordered State Space Theorem on Single Crossing and Single Peakedness Theorem Suppose the Assumption on Winning Coalitions holds. 1 If preferences are generic and satisfy single crossing and the monotonic quasi-median voter property holds, then Assumptions on Payoffs above are satisfied and Theorems 1 and 2 apply. 2 If preferences are generic and single peaked and all winning coalitions intersect (i.e., X W x and Y W y imply X Y = ), then Assumptions on Payoffs are satisfied and Theorems 1 and 2 apply. Note monotonic median voter property is weaker than the assumption that X W x Y W y = X Y =. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

75 Constitutions Examples and Applications Voting in Clubs N individuals, I = {1,..., N} N states (clubs), s k = {1,..., k} Assume single-crossing condition for all l > k and j > i, w j (s l ) w j (s k ) > w i (s l ) w i (s k ) Assume genericity : for all l > k, w j (s l ) = w j (s k ) Then, the theorem for ordered spaces applies and shows existence of MPE in pure strategies for any majority or supermajority rule. It also provides a full characterization of these equilibria. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

76 Constitutions Examples and Applications Voting in Clubs If in addition only odd-sized clubs are allowed, unique dynamically stable state. Equilibria can easily be Pareto ineffi cient. If genericity is relaxed, so that w j (s l ) = w j (s k ), then the theorem for ordered spaces no longer applies, but both the axiomatic characterization and the noncooperative theorems can still be applied from first principles. Also can be extended to more general pickle structures (e.g., weighted voting or supermajority) and general structure of clubs (e.g., clubs on the form {k n,..., k,..., k + n} I for a fixed n and different values of k). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

77 Constitutions Examples and Applications An Example of Elite Clubs Specific example: suppose that preferences are such that w j (s n ) > w j (s n ) > w j (s k ) = w j (s k ) for all n > n j and k, k < j individuals always prefer to be part of the club individuals always prefer smaller clubs. Winning coalitions need to have a strict majority (e.g., two out of three, three out of four etc.). Then, {1} is a stable club (no wish to expand) {1, 2} is a stable club (no wish to expand and no majority to contract) {1, 2, 3} is not a stable club (3 can be eliminated) {1, 2, 3, 4} is a stable club More generally, clubs of size 2 k for k = 0, 1,... are stable. Starting with the club of size n, the equilibrium involves the largest club of size 2 k n. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

78 Constitutions Examples and Applications Stable Constitutions N individuals, I = {1,..., N} In period 2, they decide whether to implement a reform (a votes are needed) a is determined in period 1 Two cases: Voting rule a: stable if in period 1 no other rule is supported by a voters Constitution (a, b): stable if in period 1 no other constitution is supported by b voters Preferences over reforms translate into preferences over a Barbera and Jackson assume a structure where these preferences are single-crossing and single-peaked Motivated by this, let us assume that they are strictly single-crossing Stable voting rules correspond to myopically (and dynamically) stable states Stable constitutions correspond to dynamically stable states Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

79 Political Eliminations Constitutions Examples and Applications The characterization results apply even when states do not form an ordered set. Set of states S coincides with set of coalitions C Each agent i I is endowed with political influence γ i Payoffs are given by proportional rule { γi /γ w i (X ) = X if i X 0 if i / X where γ X = γ j j X and X is the ruling coalition. this payoff function can be generalized to any function where payoffs are increasing in relative power of the individual in the ruling coalition Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

80 Constitutions Political Eliminations (continued) Examples and Applications Winning coalitions are determined by weighted (super)majority rule α [1/2, 1) W X = Genericity: γ X = γ Y only if X = Y {Y : j Y X γ j > α j X γ j } Assumption on Payoffs is satisfied and the axiomatic characterization applies exactly. If players who are not part of the ruling coalition have a slight preference for larger ruling coalitions, then Stronger Acyclicity Assumption is also satisfied. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

81 Constitutions Examples and Applications Other Examples Ineffi cient inertia The role of the middle class in democratization Coalition formation in democratic systems Commitment, (civil or international) conflict and peace Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

82 General Analysis Dynamic Game We now return to the general model with stochastic elements and discount factor < 1 Focus on Markov Voting Equilibrium. Comparison with Markov Perfect Equilibria again similar (and discussed below). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

83 Dynamic Game General Analysis 1 Period t begins with state s t 1 and environment E t 1 inherited from the previous period (where s 0 is exogenously given). 2 Shocks are realized. 3 Players become agenda-setters, one at a time, according to the protocol π s t 1. Agenda-setter i proposes an alternative state a t,i 4 Players vote sequentially over the proposal a t,i. If the set of players that support the transition is a winning coalition, then s t = a t,i. Otherwise, the next person makes the proposal, and if the last agent in the protocol has already done so, then s t = a t,i. 5 Each player i gets instantaneous utility w Et,i (s t ) c Et,i (s t 1, s t ). Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

84 APPROACH General Analysis Equilibria Recall that any MPE in pure strategies can be represented by a set of transition mappings {φ E } such that if s t 1 = s, and E t = E, then s t = φ E (s) along the equilibrium path, where recall that φ : S S. Transition mapping φ = {φ E : S S} is monotone if for any s 1, s 2 S with s 1 s 2, φ E (s 1 ) φ E (s 2 ). natural, given monotonic median voter property Key steps in analysis fix E characterize φ E and expected payoffs when there is no stochasticity then backward induction and dynamic programming. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

85 General Analysis Equilibria General Result Theorem (existence) There exists a Markov voting equilibrium with monotone transition mapping φ. Theorem (uniqueness) Generically there exists no other Markov voting equilibrium with monotone transition mapping if either every set of quasi-median voters is a singleton or preferences are single-peaked (plus additional conditions on transition costs; e.g., only one step transitions). Thus monotone transition mappings arise naturally. though equilibria without such monotonicity may exist. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

86 Non-uniqueness General Analysis Equilibria Political rule: unanimity in state 1 and the green player dictator in states 2 and 3. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

87 General Analysis General Characterization Limiting states and effi ciency Back to the general model: Theorem (limit behavior) In any Markov voting equilibrium, there is convergence to a limiting state with probability 1. The limiting state depends on the timing of shocks. Theorem (effi ciency) If each β E is suffi ciently small, then the limiting state is Pareto effi cient. Otherwise the limiting state may be Pareto ineffi cient. Example of Pareto ineffi ciency: elite E, middle class M E rules in a, M rules in c and d. w E (d) < w E (a) < w E (c) w M (a) < w M (c) < w M (d) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

88 General Analysis Comparative Statics Comparative statics Theorem ( monotone comparative statics) Suppose that environments E 1 and E 2 coincide on S = [1, s] S and β E 1 = β E 2, φ 1 and φ 2 are MVE in these environments. Suppose x S is such that φ 1 (x) = x. Then φ 2 (x) x. Implication, suppose that φ 1 (x) = x is reached before there is a switch to E 2. Then for all subsequent t, s t x. Intuition: if some part of the state space is unaffected by shocks, it is either reached without shocks or not reached at all. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

89 General Analysis Comparative Statics MPE vs. Markov voting equilibria Theorem (MPE MVE) For any MVE φ (monotone or not) there exists a set of protocols such that there exists a Markov Perfect equilibrium of the game above which implements φ. Conversely, if for some set of protocols and some MPE σ, the corresponding transition mapping φ = {φ E } E E is monotone, then it is MVE. In addition, if the set of quasi-median voters in two different states have either none or one individual in common, and only one-step transitions are possible, every MPE corresponds to a monotone MVE (under any protocol). For each Markov voting equilibrium, there exists a protocol π such that the resulting (pure-strategy) MPE induces transitions that coincide with the Markov voting equilibrium. Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

90 General Analysis Applications Simple example Suppose three groups: elite, middle class and workers. The elite rule under absolutist monarchy, a. Suppose that with limited franchise, c, the middle class rules with probability p and workers rule with probability 1 p. Workers rule in full democracy, d. The middle-class prefer limited franchise, workers prefer full democracy. Payoffs w E (d) < w E (a) < w E (c) w M (a) < w M (d) < w M (c) w W (a) < w W (c) < w W (d) Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

91 General Analysis Simple example (continued) Applications autocracy? limited franchise full democracy What happens if β large and p small? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

92 General Analysis Simple example (continued) Applications autocracy! limited franchise full democracy Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

93 General Analysis Simple example (continued) Applications What happens if p = 1 or close to 1? autocracy! limited franchise full democracy What happens if β small or intermediate? Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117

94 General Analysis Simple example (continued) Applications Now suppose that p takes different values in different environments. We start in E 1 and then stochastically transition to either E 2 or E 3, both of which are absorbing, and p E2 = 1 and p E3 < 1. Is an early resolution of uncertainty good for transitioning to democracy? autocracy limited franchise??? full democracy Daron Acemoglu (MIT) Political Economy Lectures 3-5 March 8, 13 and / 117